Brainstack Skills and Harness Context

Two kinds of skills

Brainstack has public product skills and shared-brain imported skills.

Public product skills ship with Brainstack. They explain how to use, debug, operate, curate, and repair Brainstack itself. These are generic and should not contain private topology.

Shared imported skills come from your machine, repo, or URL. Brainstack packages them as shared-brain imports so other connected harnesses can refresh them locally.

Install public Brainstack skills

brainctl skills install --target codex --profile client
brainctl skills install --target codex --profile operator

Use client for normal machines and operator for machines that run control-host, curation, worker, and recovery tasks.

Import shared skills

Start with a dry run when importing a whole machine's skills:

brainctl import skills \
  --config ~/.config/brainstack/brainstack.yaml

Apply only after reviewing the plan:

brainctl import skills \
  --config ~/.config/brainstack/brainstack.yaml \
  --apply

Import one skill from a local path or URL:

brainctl import skill ~/.codex/skills/brainstack/SKILL.md \
  --config ~/.config/brainstack/brainstack.yaml

brainctl import skill https://github.com/example/skill-repo \
  --config ~/.config/brainstack/brainstack.yaml

brainctl import skills scans current and default harness skill locations, prints a deterministic plan, and only writes when you add --apply.

Hooks and daemon

Hooks are fail-open convenience. They should never block a prompt just because Brainstack is offline.

brainctl hooks install --target all --config ~/.config/brainstack/brainstack.yaml
brainctl daemon install --config ~/.config/brainstack/brainstack.yaml --start

The daemon keeps the local shared-brain clone and shared skills fresh in the background, so hooks do not need to pull git before every prompt.